File size: 1,723 Bytes
3978e51 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 |
### Dataset types for training
* **Type 1 (MUSDB)**: different folders. Each folder contains all needed stems in format _< stem name >.wav_. The same as in MUSDBHQ18 dataset. In latest code releases it's possible to use `flac` instead of `wav`.
Example:
```
--- Song 1:
------ vocals.wav
------ bass.wav
------ drums.wav
------ other.wav
--- Song 2:
------ vocals.wav
------ bass.wav
------ drums.wav
------ other.wav
--- Song 3:
...........
```
* **Type 2 (Stems)**: each folder is "stem name". Folder contains wav files which consists only of required stem.
```
--- vocals:
------ vocals_1.wav
------ vocals_2.wav
------ vocals_3.wav
------ vocals_4.wav
------ ...
--- bass:
------ bass_1.wav
------ bass_2.wav
------ bass_3.wav
------ bass_4.wav
------ ...
...........
```
* **Type 3 (CSV file)**:
You can provide CSV-file (or list of CSV-files) with following structure:
```
instrum,path
vocals,/path/to/dataset/vocals_1.wav
vocals,/path/to/dataset2/vocals_v2.wav
vocals,/path/to/dataset3/vocals_some.wav
...
drums,/path/to/dataset/drums_good.wav
...
```
* **Type 4 (MUSDB Aligned)**:
The same as Type 1, but during training all instruments will be from the same position of song.
### Dataset for validation
* The validation dataset must be the same structure as type 1 datasets (regardless of what type of dataset you're using for training), but also each folder must include `mixture.wav` for each song. `mixture.wav` - is the sum of all stems for song.
Example:
```
--- Song 1:
------ vocals.wav
------ bass.wav
------ drums.wav
------ other.wav
------ mixture.wav
--- Song 2:
------ vocals.wav
------ bass.wav
------ drums.wav
------ other.wav
------ mixture.wav
--- Song 3:
...........
```
|